Vercel CEO reveals his Agentic Engineering Workflow

In this extensive interview, Vercel CEO Guillermo Rauch details his approach to 'agentic engineering,' an evolving workflow where AI agents assist throughout the software development lifecycle. Rauch explores the shift from human-centric to AI-assisted coding, emphasizing the necessity for engineers to adapt by utilizing sandboxed environments and robust testing frameworks. He argues that developers must maintain 'creative control' while offloading mundane tasks like boilerplate generation and basic debugging to AI, stressing that human judgment remains paramount in architectural and product decisions.

The conversation delves deeply into the practical implementation of AI in the development stack. Rauch outlines the role of tools like Vercel's 'AI Gateway' and 'Deepsec' in managing security risks, which he characterizes as an immeasurable 'security debt' inherent in large codebases. He emphasizes the importance of 'verification engineering,' wherein developers rigorously test the AI's output, maintaining that the future of engineering lies in composing secure, testable components rather than simply writing raw code.

Throughout the discussion, Rauch challenges conventional wisdom regarding software development speed and reliability. He advocates for a philosophy of 'iteration velocity' over raw speed, warning that shipping code too quickly without proper testing creates long-term technical and security liabilities. By framing AI as a collaborator rather than a replacement, he offers a pragmatic roadmap for engineers aiming to leverage generative AI while maintaining the rigor required for production-grade software.